Anthropic Expands Mythos AI Access to European and UK Banks as Security Concerns Mount

Gate News message, April 22 — Anthropic is expanding access to its Mythos AI model to European and UK banks, following an initial rollout to select U.S. financial institutions. The expansion will include security checks, according to Reuters. JPMorgan Chase is the only bank Anthropic has publicly confirmed as having access, though Bank of America has also tested the model internally.

Mythos operates within Project Glasswing, a program designed to protect critical software by providing early access to Claude Mythos Preview. Early partners include JPMorgan Chase, Apple, Amazon Web Services, Google, and Microsoft. The model has identified thousands of zero-day vulnerabilities, including a 27-year-old bug in OpenBSD, a security-focused open-source operating system.

U.S. Treasury and Federal Reserve officials have urged bank executives to treat Mythos as a legitimate security tool and use it to identify flaws in their own systems. However, regulators and policymakers have raised concerns about the model’s risks to banks and legacy systems.

The limited rollout has sparked debate over market fairness. Joachim Nagel, president of the Deutsche Bundesbank, argued that broader access is needed to ensure fair competition across European lenders. Market concerns also surfaced as cybersecurity vendors’ shares declined: CrowdStrike and Palo Alto Networks dropped 5% to 11% following the model’s announcement.

Mythos represents an active test case for dual-use AI systems, which can serve both defensive and offensive purposes. Resistance to the controlled release may influence how future dual-use AI models reach the market, potentially driving demand for stricter access rules and external oversight.